The engine behind every number.
ARCNM turns 3D CAD and 2D drawings into a defensible, auditable cost. Three layers do the work — deep feature extraction, manufacturing-process analysis, and cost calculation — with physics at the core and machine-learning calibration per source.
1. Deep feature extraction & PMI fusion
A B-rep geometry engine reads the 3D solid; vision-language models read the 2D drawing. The two are fused so every dimension, tolerance and GD&T callout binds to the geometry it constrains, with deterministic, re-derivable feature IDs.
2. Manufacturing-process physics
The part is classified to a discipline and planned into operations. Physics handlers (drilling, milling, turning, press brake, laser, waterjet, wire EDM, grinding, casting, moulding) compute setup and cycle time from industry-standard kernels.
3. Lot-size cost curve
Cost is a curve over batch size, not a single number. ARCNM returns unit cost at every quantity plus the economically meaningful break-points: setup/run crossover, fixture amortisation, tool life, EOQ optimum, machine-class crossover and learning-curve kinks.
Per-source calibration
A six-tier parameter cascade resolves every physics input. Alignment to a customer's real costs happens through machine-learning calibration — a high-dimensional regression at the cost posterior, with calibrated (conformal) uncertainty — while the extractor, classifier, planner and physics stay industry-standard and untouched.
Your data
Quotes are auditable end-to-end with full provenance, multi-tenant isolation, and per-tenant environments for machines, rates and materials.